Input
Waiting for input.
Output
JSON Lines is how logs, data exports and LLM fine-tuning datasets travel: one JSON document per line, appendable and streamable. It is also awkward to inspect, because no ordinary JSON tool will parse the file as a whole. This converts a .jsonl file into a proper array you can read, search and feed to the other tools here, and flattens an array back to one record per line.
Built for the NDJSON that log exports and LLM fine-tuning files actually ship as.
One corrupt record in fifty thousand is named, not hidden behind a generic failure.
Convert everything that parses and get a count of what was skipped.
Array to one-record-per-line, or a .jsonl file into an array you can actually read.
Practically none — both mean one JSON value per line, separated by newlines. The names are used interchangeably, and this tool reads and writes both.
Yes. By default it refuses and tells you which lines are bad; tick "Skip invalid lines" and it converts everything else, reporting how many it skipped.
Yes — those are JSONL. Convert to an array to inspect or edit the set, then convert back to JSONL for upload.